Hidden Markov Model(HMM)

This repo is a toy python implementation of A Revealing Introduction to Hidden Markov Model. This implementation of HMM uses scaled alpha and beta to avoid numerical underflow resulting from repeatitive multiplication of probabilities. For a case of multiple sequences being observed, multiple sequence fitting is also implemented as shown in baum-welch wikipedia

Notes

• This repo was initially developed for a Knowledge Tracing task, so the number of observation types and hidden state types are set to 2. You can easily adjust these to arbitrary integers with minor fixes.
• config.py is a setting for initial parameters (transition matrix, emission matrix, initial state probability) though it is described in knowledge tracing terms.
• Run with python HMM.py